home | download | howto | module docs | git

loss: compute a loss value for fitting purposes

this modules serves to compute a loss function to guide non-linear optimisation of module parameters. to this end, it reads an input buffer with some processing applied to it. this is then compared against the reference image. after computing the loss, it will be stored (cpu side) in a parameter loss.

connectors

• input : the processed image
• orig : the reference image
• dspy a display channel to output the masked pixels that are considered during loss computation

parameters

• bound : only pixels within these bounds are considered in the loss. pixel brightness has to be > bound.x, < bound.y and saturation (i.e. max channel minus min channel) has to be < bound.z.
• loss : the computed loss will be stored here if download sink was enabled during graph processing.